School of Nursing, Duke University, Durham, NC, USA; Center for the Study of Aging and Human Development, Duke University, Durham, NC, USA.
Department of Computer Science, Duke University, Durham, NC, USA.
J Am Med Dir Assoc. 2021 Feb;22(2):291-296. doi: 10.1016/j.jamda.2020.09.025. Epub 2020 Oct 29.
To evaluate a machine learning model designed to predict mortality for Medicare beneficiaries aged >65 years treated for hip fracture in Inpatient Rehabilitation Facilities (IRFs).
Retrospective design/cohort analysis of Centers for Medicare & Medicaid Services Inpatient Rehabilitation Facility-Patient Assessment Instrument data.
A total of 17,140 persons admitted to Medicare-certified IRFs in 2015 following hospitalization for hip fracture.
Patient characteristics include sociodemographic (age, gender, race, and social support) and clinical factors (functional status at admission, chronic conditions) and IRF length of stay. Outcomes were 30-day and 1-year all-cause mortality. We trained and evaluated 2 classification models, logistic regression and a multilayer perceptron (MLP), to predict the probability of 30-day and 1-year mortality and evaluated the calibration, discrimination, and precision of the models.
For 30-day mortality, MLP performed well [acc = 0.74, area under the receiver operating characteristic curve (AUROC) = 0.76, avg prec = 0.10, slope = 1.14] as did logistic regression (acc = 0.78, AUROC = 0.76, avg prec = 0.09, slope = 1.20). For 1-year mortality, the performances were similar for both MLP (acc = 0.68, AUROC = 0.75, avg prec = 0.32, slope = 0.96) and logistic regression (acc = 0.68, AUROC = 0.75, avg prec = 0.32, slope = 0.95).
A scoring system based on logistic regression may be more feasible to run in current electronic medical records. But MLP models may reduce cognitive burden and increase ability to calibrate to local data, yielding clinical specificity in mortality prediction so that palliative care resources may be allocated more effectively.
评估一种机器学习模型，用于预测在住院康复机构（IRF）接受髋部骨折治疗的 65 岁以上 Medicare 受益人的死亡率。
对医疗保险和医疗补助服务中心住院康复机构患者评估工具数据进行回顾性设计/队列分析。
共有 17140 人在 2015 年因髋部骨折住院后入住 Medicare 认证的 IRF。
患者特征包括社会人口统计学（年龄、性别、种族和社会支持）和临床因素（入院时的功能状态、慢性疾病）和 IRF 住院时间。结果是 30 天和 1 年全因死亡率。我们训练和评估了 2 种分类模型，逻辑回归和多层感知器（MLP），以预测 30 天和 1 年死亡率，并评估了模型的校准、区分和精度。
对于 30 天死亡率，MLP 表现良好[准确度（acc）=0.74，接受者操作特征曲线下的面积（AUROC）=0.76，平均精度（avg prec）=0.10，斜率=1.14]，逻辑回归也表现良好（acc=0.78，AUROC=0.76，avg prec=0.09，斜率=1.20）。对于 1 年死亡率，MLP（acc=0.68，AUROC=0.75，avg prec=0.32，斜率=0.96）和逻辑回归（acc=0.68，AUROC=0.75，avg prec=0.32，斜率=0.95）的表现相似。
基于逻辑回归的评分系统可能更适合在当前的电子病历中运行。但是，MLP 模型可以减少认知负担并提高校准到本地数据的能力，从而在死亡率预测中产生临床特异性，以便更有效地分配姑息治疗资源。
